Frank van der Linden writes:

| After all, INET6 is scheduled to be the default in the future.

This is wildly optimistic.

| We should assume whatever's in GENERIC.

Then why doesn't sendmail(8) fail if I do not have "options ISO",
and have not made a configuration change?

Why doesn't sendmail(8) talk OSI, XNS, or Appletalk anyway?
At least one of those is an International Standard with as
much weight as, and even greater deployment than, IPv6.

| If you go through the trouble of
| editing out the INET6 option of your config file, it's not unreasonable
| to expect that you might have to edit a few more config files.
| What should not happen is binaries failing without having the option to
| tell them about your non-v6 kernel via a config file, of course.

Uh, I don't like the idea of having to scour userland for undocumented
configuration needs if I comment out GENERIC device drivers, busses,
protocols, pseudo-devices, DEBUG options, etc. etc. etc.  

I think itojun has done a pretty good job so far of experimenting
with v6 in userland in a generally non-disruptive way.   To the extent
that his pragmatic "one userland" approach to things works, I'm grateful.

This particular (certainly unintentional) failure was a small
deviation from that.  As a result I no longer use sendmail.    
I now use postfix, which doesn't care configuration-wise whether I
am running an IPv4 kernel or a GENERIC that has both IPv4 and IPv6
support, and am also grateful to Perry and others for bringing it
into the tree, since I like it better than bendmail so far.

My personal desire for the ability to easily prepare a system wholly
devoid of ANY v6-related data outside of the source tree you can consider
a whine, since my contribution to NetBSD is rather small.

	Sean.